Hi Guys! This is my first post after lurking for a 4 days or so...
I've got some efficency (sp?) issues after building a SMPS. I started building it on monday and I've now tested it and Its taking about 250W (14V @ 18A) input and only putting out 154W (+-34V @ 2.27A) . Its losing about 100W internally. I've used 16*0.4mm2 for the primary at 4 turns and 5*0.4mm2 at 11 turns for the secondary.
The basic layout of the circuit is:
2*470uF caps on the input rails
2*4 turns primary. Center attached to 14V, outside attached to ground alternately. Each side is attached for 37.5% of the time. (This minimises both the 3rd and 5th harmonics)
The core is a ETD49 with 3c90 material
The secondary is 2*11 turns rectified into 4*470uF for each rail, center tapped.
Mosfets for primary: irf3205*2 for each side of primary. They don't even get warm. Turned on/off in about 0.5uS.
Rectification diodes on secondary: BYV28-200 diodes (3.5A, 200V diodes with 25ns turn-off time and 200v reverse voltage). No hotter than can be expected (can touch with fingers without burning sensation setting in at extended time periods).
So can you guys help me out as to where I might be lsing all my power? Faulty meters (I used different meters for the current measurement but not the voltage) or lots of power going into the copper windings? I'm stumped myself - I ran it for 10 minutes today at 250W going in and it was still ok - the output voltage dropped from 68v across both output rails to 67.8v in that time peroid.
Cheers in advance for any helpful hints/pointing out stoopid things I might of done.
I've got some efficency (sp?) issues after building a SMPS. I started building it on monday and I've now tested it and Its taking about 250W (14V @ 18A) input and only putting out 154W (+-34V @ 2.27A) . Its losing about 100W internally. I've used 16*0.4mm2 for the primary at 4 turns and 5*0.4mm2 at 11 turns for the secondary.
The basic layout of the circuit is:
2*470uF caps on the input rails
2*4 turns primary. Center attached to 14V, outside attached to ground alternately. Each side is attached for 37.5% of the time. (This minimises both the 3rd and 5th harmonics)
The core is a ETD49 with 3c90 material
The secondary is 2*11 turns rectified into 4*470uF for each rail, center tapped.
Mosfets for primary: irf3205*2 for each side of primary. They don't even get warm. Turned on/off in about 0.5uS.
Rectification diodes on secondary: BYV28-200 diodes (3.5A, 200V diodes with 25ns turn-off time and 200v reverse voltage). No hotter than can be expected (can touch with fingers without burning sensation setting in at extended time periods).
So can you guys help me out as to where I might be lsing all my power? Faulty meters (I used different meters for the current measurement but not the voltage) or lots of power going into the copper windings? I'm stumped myself - I ran it for 10 minutes today at 250W going in and it was still ok - the output voltage dropped from 68v across both output rails to 67.8v in that time peroid.
Cheers in advance for any helpful hints/pointing out stoopid things I might of done.
smps efficiency issues
Is this a continuous or discontinuous conduction mode circuit? How large is the inductor, and what is the switching frequency? I would suspect that the primary current measurement may not be correct. If the inductor current is discontinuous, then there will be a large ac ripple current in the primary. Are you using a true rms ammeter? If nothing is getting warm, I doubt that 100 watts is being dissipated, since that is a lot of power. The efficiency computes to be 61.6 percent, low by modern day standards. Are the snubber components getting hot? How about the inductor? If you have a scope, maybe you can observe the primary current waveform using a sensing resistor. The waveform should be a small ac ripple resting on a large dc pedestal Also, power can be lost in the input and/or output filter caps, as well as the inductor. The PWM control IC dissipates the power lost in the gate drive of the MOSFETs. Also, the transformer's core and windings account for some power loss. Lastly, what is the source that is driving the input of the PWM IC? The internal resistance of the source, as well as that of the wiring produces loss. This is the hidden efficiency killer in many cases. My off the cuff guess is that the 18 amp measured value of input current should be re-examined. Best regards.
Is this a continuous or discontinuous conduction mode circuit? How large is the inductor, and what is the switching frequency? I would suspect that the primary current measurement may not be correct. If the inductor current is discontinuous, then there will be a large ac ripple current in the primary. Are you using a true rms ammeter? If nothing is getting warm, I doubt that 100 watts is being dissipated, since that is a lot of power. The efficiency computes to be 61.6 percent, low by modern day standards. Are the snubber components getting hot? How about the inductor? If you have a scope, maybe you can observe the primary current waveform using a sensing resistor. The waveform should be a small ac ripple resting on a large dc pedestal Also, power can be lost in the input and/or output filter caps, as well as the inductor. The PWM control IC dissipates the power lost in the gate drive of the MOSFETs. Also, the transformer's core and windings account for some power loss. Lastly, what is the source that is driving the input of the PWM IC? The internal resistance of the source, as well as that of the wiring produces loss. This is the hidden efficiency killer in many cases. My off the cuff guess is that the 18 amp measured value of input current should be re-examined. Best regards.
I realised that I hadn't mentioned the switching frequency when I first posted it, but since I'm new I couldn't put in changes in. Its switching at 21kHz, with no feed back. The windings do get warm, but can still touch them and the core is still pretty cool to touch.
I think they layout is ok - I've got a two sided PCB with 14V on one side and ground on the other - and similar for the output, and no leads are longer than 40mm.
Claude: Its nothing flash like a boost or cuk converter - its a just a transformer connected one way then the other to 14V, rectified on the otherside - no inductors, just smoothing caps.
As for the measurements - I suppose I'll have to find some beefy low ohm resisters and measure the voltage drop over them with a osiloscope to get a good measurement and to find out if its peak or rms measured current and do the same for the voltage.
I'll get back with the real results.
I think they layout is ok - I've got a two sided PCB with 14V on one side and ground on the other - and similar for the output, and no leads are longer than 40mm.
Claude: Its nothing flash like a boost or cuk converter - its a just a transformer connected one way then the other to 14V, rectified on the otherside - no inductors, just smoothing caps.
As for the measurements - I suppose I'll have to find some beefy low ohm resisters and measure the voltage drop over them with a osiloscope to get a good measurement and to find out if its peak or rms measured current and do the same for the voltage.
I'll get back with the real results.
In order to meter the input power you need the average DC-voltage times the average DC-current.
If your input voltage is instable then things are getting more complicated, but probably this is not the issue.
You can measure the average DC input current quite easy, without scope.
One current shunt resistor. Then use a R-C low pass 1kOhm-100uF.
The voltage across the 100uF cap will perfectly reflect the average DC current.
From your description about the ETD 49, I would not expect that you have excessive losses there.
If you still can touch the windings they may be around 60°C.
Core is fairly cool....
If this is the situation when the SMPS is in thermal steady state, then
you may have about 5W losses in that transformer.
If there is nothing getting really hot, then I would expect that
you do not have 100W losses.
Cheers
Markus
If your input voltage is instable then things are getting more complicated, but probably this is not the issue.
You can measure the average DC input current quite easy, without scope.
One current shunt resistor. Then use a R-C low pass 1kOhm-100uF.
The voltage across the 100uF cap will perfectly reflect the average DC current.
From your description about the ETD 49, I would not expect that you have excessive losses there.
If you still can touch the windings they may be around 60°C.
Core is fairly cool....
If this is the situation when the SMPS is in thermal steady state, then
you may have about 5W losses in that transformer.
If there is nothing getting really hot, then I would expect that
you do not have 100W losses.
Cheers
Markus
@tensop:
You may need the powerfactor if you measure the rms-values of voltage and current.
Usually people talk about the power factor in AC applications.
To "compensate" the powerfactor in a proper way may sometimes be quite complicated as the powerfactor is influenced by the the phase shift of current versus voltage (easy to compensate) AND the input current harmonics (difficult to "compensate").
But we are lucky here. We have a fairly stable DC input.
If I understand right our Flying Dutch Man will operate from a
a car battery.
So for power measurement we can work with simple equipment.
Bye
Markus
You may need the powerfactor if you measure the rms-values of voltage and current.
Usually people talk about the power factor in AC applications.
To "compensate" the powerfactor in a proper way may sometimes be quite complicated as the powerfactor is influenced by the the phase shift of current versus voltage (easy to compensate) AND the input current harmonics (difficult to "compensate").
But we are lucky here. We have a fairly stable DC input.
If I understand right our Flying Dutch Man will operate from a
a car battery.
So for power measurement we can work with simple equipment.
Bye
Markus
ChocoHolic - your right on the mark - I want to get fairly good power rails to run a stero off a car battery (while running of course, so hence the 14V). The problem was that I was using fiarly simple measurements - the DC power supply gives a needle voltage and amp reading - but I was running it in the lower end of its capababilities (can supply 48V @ 50A). But I'm pretty sure it was in thermal steady state after 10mins.
Well as for finding a chunky resistor that can handle the jandle - didn't happen (at 0.015 ohms @ 18A its still 5W). However I did manage to locate a clamp-on DC current meter that plugs into a scope - give 100mV per Amp. I didn't have time today to get read outs, but I'll get onto it and narrow down the problem. I'll even post up sexy waveforms straight from the o'scope.
If the primary inductance does prove to be the problem I'll crank up the frequency by changing a cap or a resistor... Ready to be shocked? I'm using a 555 to generate the timing and a couple of comparators to get the right duty cycle to drive the FETs (though some BJTs). It is working quite well though - they both have the same duty cycle +- 0.1% - so there shouldn't be any serious flux walking - and if there was I'd just have a heater wouldn't I?
Also for posting pictures - do I have to host them and link them in or is there somewhere I can upload them to? I can host them if that is the case - but its hard to stick stuff on and my websever isn't really the most reliable of creatures. I'll get them up here in a couple o days - I've got some
to do tomorrow.
Well as for finding a chunky resistor that can handle the jandle - didn't happen (at 0.015 ohms @ 18A its still 5W). However I did manage to locate a clamp-on DC current meter that plugs into a scope - give 100mV per Amp. I didn't have time today to get read outs, but I'll get onto it and narrow down the problem. I'll even post up sexy waveforms straight from the o'scope.
If the primary inductance does prove to be the problem I'll crank up the frequency by changing a cap or a resistor... Ready to be shocked? I'm using a 555 to generate the timing and a couple of comparators to get the right duty cycle to drive the FETs (though some BJTs). It is working quite well though - they both have the same duty cycle +- 0.1% - so there shouldn't be any serious flux walking - and if there was I'd just have a heater wouldn't I?
Also for posting pictures - do I have to host them and link them in or is there somewhere I can upload them to? I can host them if that is the case - but its hard to stick stuff on and my websever isn't really the most reliable of creatures. I'll get them up here in a couple o days - I've got some
to do tomorrow.
It if really was loosing 100 watts it would be smoking in under 10 seconds!!
You aren't concerned with harmonics here. Provided you aren't using a choke-input filter on the rectifier side, run as wide a duty cycle as you can. Should give lower peak currents on the mosfets and will be generally nicer all round.
That's as slow as you would ever want to go. Shoot for 0.1uS for low switching losses. Not a real big deal though if you have a capacitor-input filter on the output because (neglecting tranny leakage reactance) the fets will be switching off at practically ZV/ZC.
Use schottky's here if at all possible.
Last of all -> IME many digital meters go nuts if you feed high frequencies into them.
You might not be measuring anything like the correct amps.
Just in order to make sure that I am on the right track...
I estimate your SMPS is a similar topology to this:
http://sound.westhost.com/project89.htm
Right?
Then the primary inductance is not of major importance.
You are using an ungapped core right?
Your primary is 2x4turns center tapped?
Then it is fine. You will not have any issues with saturation
at 21kHz.
Swing of flux density will run between +/- 150mT.
deltaB = (U x Ton) / (N x Ae)
= (14V x 17.9µs) / (4 x 210 E-6 m^2) = 0.3T = 300mT =+/-150mT
If you increase the frequency you will probably not get a benefit, without adjusting the entire design.
...from all your description my feeling is that your circuit works
better than your measurement
I partially agree to Circlotron. Especially his last comment.
Because of this my proposal was to measure the DC values
after a simple filter.
But you do not need to try decrease the losses in components
which remain cool. If your MOSFETs are cool then they do not have much losses.
Are you sure that the turn on/off time is really 0.5µs?
Is this the speed of the voltage sloping or the current sloping???
Well, who cares. If they are cold, they are not lossy.
Good luck
Markus
I estimate your SMPS is a similar topology to this:
http://sound.westhost.com/project89.htm
Right?
Then the primary inductance is not of major importance.
You are using an ungapped core right?
Your primary is 2x4turns center tapped?
Then it is fine. You will not have any issues with saturation
at 21kHz.
Swing of flux density will run between +/- 150mT.
deltaB = (U x Ton) / (N x Ae)
= (14V x 17.9µs) / (4 x 210 E-6 m^2) = 0.3T = 300mT =+/-150mT
If you increase the frequency you will probably not get a benefit, without adjusting the entire design.
...from all your description my feeling is that your circuit works
better than your measurement
I partially agree to Circlotron. Especially his last comment.
Because of this my proposal was to measure the DC values
after a simple filter.
But you do not need to try decrease the losses in components
which remain cool. If your MOSFETs are cool then they do not have much losses.
Are you sure that the turn on/off time is really 0.5µs?
Is this the speed of the voltage sloping or the current sloping???
Well, who cares. If they are cold, they are not lossy.
Good luck
Markus
Another point, but it does not look like you are
struggling with this.
Symmetry of push/pull duty cycle very important in this
circuit. Already small unsymetries will run your transformer into saturation....
Furthermore the data sheet of your core material:
http://www.elnamagnetics.com/library/catalogs/ferroxcube/2002_handbook/mat3c90.pdf
struggling with this.
Symmetry of push/pull duty cycle very important in this
circuit. Already small unsymetries will run your transformer into saturation....
Furthermore the data sheet of your core material:
http://www.elnamagnetics.com/library/catalogs/ferroxcube/2002_handbook/mat3c90.pdf
ChocoHolic said:
...and I thought the frequency should be higher...
Well there's my intuition, I'm just used to much higher voltage when talking SMPS, should have done the math too, shame on me!
But I am not sure about your saying it's +/- 150 mT, it's in my opinion +/- 300 mT, it's push pull action.
If switching frequency is 21 kHz then we have in reality 42 kHz since upper and lower FET's are switching with 180 degree phase shift.
So the 17,9 uS you stated is actually 75% duty cycle, eg. there's 25% time left until the opposit FET is switching, right...?!
When saying 75% duty cycle I mean that 75% percent of the time window is used until the opposit FET is switching, just in case so we don't rise a semantic question what duty cycle is because at the moment I must admit i don't remember how it's defined for a half bridge topology.
At 100 degrees 300 mT is maybe a bit too much, so either should the frequency be a bit higher or duty cycle a bit lower if we will experience high temperature, which hot runing equipment might be possible to reach in a car at summer time...
The core is saturating earlier at high temperature and the current running through the switching FET's is increasing with hotter FET's as a result etc.
Cheers
I think 37.5% is the percentage of the full period.
Period is 47.6us at 21kHz. So ON-time would be
17.86us. Then there is duration of about 6us while all MOSFETs
are off. After the opposite MOSFET will turn on.
The first MOSFET was "pushing".
The second MOSFET now is "pulling" which means that the
it will pull also the flux density to the other direction.
Of course in the beginnig until steady state is reached the flux density
may not be symmetrical to zero, but shifted. After some time the
it will balance (if both cycles are really symetrical !!).
Some PWM-chips provide a soft ramp up of the duty cycle in order to
avoid saturation issues during start up.
I am just thinking little more....
May be that the 37.5% duty cycle is only the duty cycle of the current.
The voltages across the winding may act different.
Probably they will make 50/50 cycle.
Which means that our flux density would ramp between +/-200mT
in steady state.
I estimate the 50/50 duty cycle of the voltages across the windings because the following:
First MOSFET was ON and will be turned OFF:
Now all MOSFETs are OFF.
Due to the law of induction the voltage across the windings will
reverse the voltage up to any clamping voltage or defect of isolation.
In our circuit it will be clamped by the secondary windings and the
output caps, if coupled perfectly. The voltage across the switch
which was turned OFF last will jump to 2x14V=28V (I think).
The voltage accross the other MOSFET will jump to zero.
A real transformer may have some leakage inductance and cause
some voltage peaks at the MOSFETs. Probably some snubber
would be a good idea!
The voltage across the winding should jump from +14V to -14V.
The stored energy in the inductor will decrease, but before it reaches zero the second MOSFET will be turned ON.
Uhps, that's great!
This means turn ON is a Zero Voltage Switching!
So switching losses will only happen during turn OFF for
the switches....
Hm, and these turn OFF losses you can probably avoid by the same
snubber which you should use anyway against voltage peaks.
...start liking this circuit, I will go to simulation and revert.....
Bye
Markus
P.S.
3C90 can handle slightly more than 300mT at 100°C without
saturation.
Period is 47.6us at 21kHz. So ON-time would be
17.86us. Then there is duration of about 6us while all MOSFETs
are off. After the opposite MOSFET will turn on.
The first MOSFET was "pushing".
The second MOSFET now is "pulling" which means that the
it will pull also the flux density to the other direction.
Of course in the beginnig until steady state is reached the flux density
may not be symmetrical to zero, but shifted. After some time the
it will balance (if both cycles are really symetrical !!).
Some PWM-chips provide a soft ramp up of the duty cycle in order to
avoid saturation issues during start up.
I am just thinking little more....
May be that the 37.5% duty cycle is only the duty cycle of the current.
The voltages across the winding may act different.
Probably they will make 50/50 cycle.
Which means that our flux density would ramp between +/-200mT
in steady state.
I estimate the 50/50 duty cycle of the voltages across the windings because the following:
First MOSFET was ON and will be turned OFF:
Now all MOSFETs are OFF.
Due to the law of induction the voltage across the windings will
reverse the voltage up to any clamping voltage or defect of isolation.
In our circuit it will be clamped by the secondary windings and the
output caps, if coupled perfectly. The voltage across the switch
which was turned OFF last will jump to 2x14V=28V (I think).
The voltage accross the other MOSFET will jump to zero.
A real transformer may have some leakage inductance and cause
some voltage peaks at the MOSFETs. Probably some snubber
would be a good idea!
The voltage across the winding should jump from +14V to -14V.
The stored energy in the inductor will decrease, but before it reaches zero the second MOSFET will be turned ON.
Uhps, that's great!
This means turn ON is a Zero Voltage Switching!
So switching losses will only happen during turn OFF for
the switches....
Hm, and these turn OFF losses you can probably avoid by the same
snubber which you should use anyway against voltage peaks.
...start liking this circuit, I will go to simulation and revert.....
Bye
Markus
P.S.
3C90 can handle slightly more than 300mT at 100°C without
saturation.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.